Create a Filter-Matrix

A Filter-Matrix can be used in combination with mutUniformMetaReset for heuristic-supported biased mutation.

makeFilterMat(
  task,
  filters,
  expectfeatfrac = 0.5,
  expectfeats = getTaskNFeats(task) * expectfeatfrac,
  minprob = 0,
  maxprob = 1
)

Arguments

task	[Task] The task to generate filter information for.
filters	[character] The filters to use. Special vilter "DUMMY" gives a constant column of expectfeatfrac.
expectfeatfrac	[numeric(1)] The expected fraction of features to have in equilibrium. Ignored if expectfeats is given.
expectfeats	[numeric(1)] The expected number of features to have in equilibrium.
minprob	[numeric(1)] The minimum probability for each feature.
maxprob	[numeric(1)] The maximum probability for each feature.

Value

matrix

Examples

library("mlr")

# Example for iris task
filters <- c("praznik_JMI", "anova.test", "variance", "DUMMY")
fima <- makeFilterMat(iris.task, filters = filters)
print(fima)
#>      value.praznik_JMI value.anova.test value.variance DUMMY
#> [1,]         0.6666667        0.3333333      0.6666667   0.5
#> [2,]         0.0000000        0.0000000      0.0000000   0.5
#> [3,]         0.3333333        1.0000000      1.0000000   0.5
#> [4,]         1.0000000        0.6666667      0.3333333   0.5